caling up the TSO by parallelizing the batched writes to the commit table. The architecture mostly changed by adding batch events and creating handlers under PersistenceProcessorImpl.
This commit increases the performance of Omid from 55K tps to 380K tps when using 6 region servers.
Please note that the addition of MonitoringContext significantly degrades performance and therefore, should becomes optional to achieve good performance.
Moved branch ohadshacham:concurrentCTwrite to omid repo. No other changes. Old PR - https://github.com/yahoo/omid/pull/75
Rebased, please comment below.
caling up the TSO by parallelizing the batched writes to the commit table. The architecture mostly changed by adding batch events and creating handlers under PersistenceProcessorImpl. This commit increases the performance of Omid from 55K tps to 380K tps when using 6 region servers.
Please note that the addition of MonitoringContext significantly degrades performance and therefore, should becomes optional to achieve good performance.